npv labs

Multimodal ML Engineer

Computer Vision +1Confirmed live in the last 24 hours

Full-time

About the Job

Requirements

At least 3 years of experience training large-scale multimodal models.
Strong PyTorch and distributed training experience, including DeepSpeed and FSDP.
Deep familiarity with multimodal architectures such as LLaVA, Qwen-VL, InternVL, Audio Flamingo, Whisper, HuBERT, or Conformer.
Hands-on reinforcement learning from human feedback and alignment experience across modalities, including GRPO, DPO, and reward modeling.
Experience with both audio and video, including sequence modeling for each, large-scale dataset curation, and production inference optimization.
Willingness to relocate to Paris or London and work in a hybrid arrangement.

Responsibilities

Train and fine-tune large-scale multimodal models for vision-language, audio, speech, and video, both from scratch and from pretrained checkpoints.
Design experiments, build multimodal data pipelines, and train mixture-of-experts architectures.
Build alignment pipelines using supervised fine-tuning, direct preference optimization, and group relative policy optimization; optimize models for production through quantization, distillation, and streaming; and deploy end-to-end.
Define evaluation metrics relevant to the product.

Desired Qualifications

Audio signal processing fundamentals, including spectrograms, mel features, and noise reduction.
Experience with mixture-of-experts architectures.

We're looking for a Multimodal ML Engineer to join White Circle, an AI Safety company building the safety, reliability, and optimization layer for AI systems through natural-language policies it automatically tests, enforces, and improves at scale. Backed by $70M (Series A) from top funds and senior leaders at OpenAI, Anthropic, HuggingFace, Mistral, DeepMind, and others, White Circle processes 100M+ API calls monthly and fine-tunes and trains its own LLMs to run faster and cheaper than open or proprietary models.

You will

Train and fine-tune large-scale multimodal models (vision-language, audio, speech, video) from scratch and from pretrained checkpoints.
Design experiments, build multimodal data pipelines, and train MoE architectures.
Build alignment pipelines (SFT, DPO, GRPO), optimize for production (quantization, distillation, streaming), and deploy end-to-end.
Define evaluation metrics that actually matter for the product.

Requirements

3+ years training large-scale multimodal models.
Strong PyTorch and distributed training experience (DeepSpeed, FSDP).
Deep familiarity with multimodal architectures – LLaVA, Qwen-VL, InternVL, Audio Flamingo, Whisper, HuBERT, Conformer or similar.
Hands-on RLHF/alignment across modalities (GRPO, DPO, reward modeling).
Both audio and video experience required – sequence modeling for each, plus large-scale dataset curation and production inference optimization.
Relocation to Paris or London (hybrid) required.

Bonus

Audio signal processing fundamentals – spectrograms, mel features, noise reduction.
MoE architecture experience.

We offer

Competitive salary + equity.
Official employment, visa and relocation help.

About the Company

npv labs

Company size: Not listed
Company stage: Not listed
Total funding: Not listed
Headquarters: Not listed
Founded: Not listed

Recently Posted Jobs

Explore more opportunities at npv labs

See more jobs